Research On Hidden Danger Prediction Method Of Logistics Enterprises Based On Graph Autoencoder

After more than 30 years of development,the logistics industry has become a pillar industry of the national economy and an important modern service industry.The development of networking,automation and unmanned technology has enabled the logistics industry to make great progress in improving distribution efficiency and reducing distribution costs.However,the working environment of most logistics personnel has not changed much compared with the past,and all kinds of potential accidents occur frequently.Due to the characteristics of logistics enterprises involved in transportation and warehousing,once accidents happen,the impact is wide and all kinds of losses are large.Accidents are caused by the accumulation of hidden dangers.How to improve the efficiency and accuracy of hidden danger investigation is an urgent problem to be solved.The previous prediction methods of hidden danger are often aimed at the occurrence probability of one kind of hidden danger or the overall number of hidden dangers,but not at the occurrence probability of any type of hidden danger.The purpose of this thesis is to predict the occurrence probability of common hidden dangers in logistics enterprises in the long run.It is divided into two parts.The first is to solve the problem of unclear class spacing in the existing hidden danger classification of logistics enterprises.Use TextCNN to embed the short text of the hidden danger description with sufficient information in the semantic space,and then use the K-means algorithm to cluster to ensure that the distance between the hidden danger classes after classification is clear,which improves the accuracy of subsequent predictions.After that,the problem of predicting the hidden dangers of the enterprise was transformed into the scoring the possibility of occurrence of hidden dangers.Use the improved network structure based on the graph autoencoder to separately encode hidden dangers and enterprises,integrate the hidden danger related information and the occurrence of hidden dangers of the enterprise,and finally use the bilinear decoder to complete the completion of the score.The data on hidden dangers of logistics companies in Beijing from 2018 to 2019 was used to verify the feasibility and accuracy of this method.Compared with the original GCMC algorithm,the error is reduced by 5.03%...